Sorting Algorithm Problem

Bentley iTwin services receive model update records that must be processed in a deterministic order. Implement a stable merge sort to order updates by priority, timestamp, and element identifier.

Each update is represented as a dictionary with keys element_id, priority, and timestamp. Return a new list without modifying the input. Sort using these rules, in order:

1. Higher priority values come first.
2. For equal priorities, earlier timestamp values come first.
3. For equal priority and timestamp, lexicographically smaller element_id values come first.
4. If all sort fields are equal, preserve the original input order.

Do not call Python's built-in sorted() or .sort(). Your implementation should use merge sort and may define helper functions.

Formal Specification

Implement sort_model_updates(updates), where updates is a list of dictionaries. Return a new list containing the same dictionaries in the required order. Timestamps are integers, and priorities are integers.